ai will end discrimination. - hiig
TRANSCRIPT
AI will end discrimination.
As part of society, AI is deeply rooted in it and as such not separable from structures of discrimination. Due to this socio-technical embeddedness, AI cannotmake discrimination disappear by itself .
WHY AI – UNRAVELING 15 MYTHS ABOUT AUTOMATION, ALGORITHMS, SOCIETY AND OURSELVES4
Artificial Intelligence ends discrimination, doesn‘t it?
Hmm. Unlike me,
other people are
so emotional,
moody, unfair,
fallible.
Instead, Artificial Intelligence seems to make better choices
than people.
Well, at least, Artificial Intelligence does not judge people by their
status... …because Artificial Intelligence can be
instructed in a way to treat all equal
people/groups equally.
Thus, can AI end discrimination in
the world?
Source images: https://www.britannica.com/topic/The-Thinker-sculpture-by-Rodin
AI: associations
correct
has no agency
objective
neutral
independent
nonjudgmental
unbiased/
fair
mathematical
autonomous
unemotional
…
Source Image: https://www.plm.automation.siemens.com/global/en/webinar/artificial-intelligence-in-automotive-drives-performance-engineering/91273
AI: associations
correct
has no agency
objective
neutral
independent
nonjudgmentalunbiased/
fair
mathematical
autonomous
unemotional
…
?Source Images: left: Photo by Owen Beard on Unsplash // right Photo by Izabela Rutkowski on Unsplash
AI: associations
correct
has no agency
objective
neutral
independent
nonjudgmentalunbiased/
fair
mathematical
autonomous
unemotional
…
!Source Images: left: Photo by Owen Beard on Unsplash // right Photo by Izabela Rutkowski on Unsplash
What is AI?
„Intelligence demonstrated by machines“ (Wikipedia)
● Originally: A field of study
● Today: Anything between science-fiction
dys-/utopias and technologies exhibiting reasoning
● Here: Data-driven (semi-)automatic decision-making
systems
Zakharov, Egor, et al. "Few-Shot Adversarial Learning of Realistic Neural Talking Head Models.“ (2019)
AI gone wrong
Source Images: https://towardsdatascience.com/algorithm-bias-in-artificial-intelligence-needs-to-be-discussed-and-addressed-8d369d675a70?gi=6da14f8698f3 https://futurezone.at/netzpolitik/der-ams-algorithmus-ist-ein-paradebeispiel-fuer-diskriminierung/400147421
WHY AI – UNRAVELING 15 MYTHS ABOUT AUTOMATION, ALGORITHMS, SOCIETY AND OURSELVES10
AI „in the wild“
Aha! AI “in the wild”, like outside a laboratory environment or outside
of some programmer geniuses’ garages is pretty discriminatory,
sometimes even worse than humans.
Source images: https://www.britannica.com/topic/The-Thinker-sculpture-by-Rodin
AI as sociotechnical system
(computer) Science
modeldata
AI practitioner
society
research teams
history
private
companies
business model
Source Image: https://www.plm.automation.siemens.com/global/en/webinar/artificial-intelligence-in-automotive-drives-performance-engineering/91273
AI as sociotechnical system
(computer) Science
modeldata
AI practitioner
society
research teams
history
private
companies
business model
Source Image: https://www.plm.automation.siemens.com/global/en/webinar/artificial-intelligence-in-automotive-drives-performance-engineering/91273
AI as sociotechnical system
(computer) Science
modeldata
AI practitioner
society
research teams
history
private
companies
business model
Source Image: https://www.plm.automation.siemens.com/global/en/webinar/artificial-intelligence-in-automotive-drives-performance-engineering/91273
Representation issues
Source: https://ars.electronica.art/outofthebox/de/gender-shades/
AI as sociotechnical system
(computer) Science
modeldata
AI practitioner
society
research teams
history
private
companies
business model
Source Image: https://www.plm.automation.siemens.com/global/en/webinar/artificial-intelligence-in-automotive-drives-performance-engineering/91273
„Data fundamentalism“
Source:Michal Kosinski and Yilun Wang. «Deep neural networks are more accurate than humans at detecting sexual orientation from facial images » By the New York Times
AI as sociotechnical system
(computer) Science
modeldata
AI practitioner
society
research teams
history
private
companies
business model
Source Image: https://www.plm.automation.siemens.com/global/en/webinar/artificial-intelligence-in-automotive-drives-performance-engineering/91273
„Data fundamentalism“
Source: Faception Website in 2018 https://www.faception.com/our-technology
AI as sociotechnical system
(computer) Science
modeldata
AI practitioner
society
research teams
history
private
companies
business model
Source Image: https://www.plm.automation.siemens.com/global/en/webinar/artificial-intelligence-in-automotive-drives-performance-engineering/91273
AI as sociotechnical system
(computer) Science
modeldata
AI practitioner
society
research teams
history
private
companies
business model
Source Image: https://www.plm.automation.siemens.com/global/en/webinar/artificial-intelligence-in-automotive-drives-performance-engineering/91273
Source: What is standpoint theory? http://hennessy.iat.sfu.ca/wp/stc2018/2018/03/04/what-is-standpoint-theory/
AI as sociotechnical system
(computer) Science
modeldata
AI practitioner
society
research teams
history
private
companies
business model
Source Image: https://www.plm.automation.siemens.com/global/en/webinar/artificial-intelligence-in-automotive-drives-performance-engineering/91273
AI as sociotechnical system
(computer) Science
modeldata
AI practitioner
society
research teams
history
private
companies
business model
Source Image: https://www.plm.automation.siemens.com/global/en/webinar/artificial-intelligence-in-automotive-drives-performance-engineering/91273
AI as sociotechnical system
(computer) Science
modeldata
AI practitioner
society
research teams
history
private
companies
business model
Source Image: https://www.plm.automation.siemens.com/global/en/webinar/artificial-intelligence-in-automotive-drives-performance-engineering/91273
Source: Frontexhttps://frontex.europa.eu/media-centre/news/news-release/artificial-intelligence-based-capabilities-for-european-border-and-coast-guard-1Dczge
What now? Emancipatory AI or burn it?
AI...• … is not separable from the
socio-political• … will not fix discrimination• … encodes social inequalities
Source Images: https://www.plm.automation.siemens.com/global/en/webinar/artificial-intelligence-in-automotive-drives-performance-engineering/91273 https://tenor.com/view/calvin-and-hobbes-calvin-fly-clouds-birds-gif-4953836 https://i.pinimg.com/originals/4d/ca/d9/4dcad98ff2d9aba671b56957ab5d70a2.gif
What now? Emancipatory AI or burn it?
“I argue that tech fixes often hide, speed up, and even deepen discrimination, while appearing to be neutral or benevolent when compared to the racism of a previous era. This set of practices that I call the New Jim Code encompasses a range of discriminatory designs – some that explicitly work to amplify hierarchies, many that ignore and thus replicate social divisions, and a number that aim to fix racial bias but end up doing the opposite” (Ruha Benjamin, Race after Technology, 2020, p. 9)
Source Images: https://www.plm.automation.siemens.com/global/en/webinar/artificial-intelligence-in-automotive-drives-performance-engineering/91273 https://tenor.com/view/calvin-and-hobbes-calvin-fly-clouds-birds-gif-4953836
What now? Emancipatory AI or burn it?
AI...• … needs a socio-technical
decoding • … may serve as a proxy to discuss
deeply rooted societal inequalities
Source Images: https://www.plm.automation.siemens.com/global/en/webinar/artificial-intelligence-in-automotive-drives-performance-engineering/91273 https://tenor.com/view/calvin-and-hobbes-calvin-fly-clouds-birds-gif-4953836 https://i.pinimg.com/originals/4d/ca/d9/4dcad98ff2d9aba671b56957ab5d70a2.gif
Why discrimination cannot just end
o Discrimination is no “irrational” wrongdoing based on individual choices whether these choices are made by an individual or an AI.
o Discrimination is a structural, systematic phenomena. Different modes of discrimination like class, racism, sexism, antisemitism, patriarchy cannot be abolished or altered by individual (conscious or unconscious) choices.
o Understanding AI from an intersectional sociotechnical perspective means: A ‚fair‘ AI can treat equal individuals or groups equally, but it still is embedded into wider social structures of discrimination and oppression and does not imply a certain treatment is just.
WHY AI – UNRAVELING 15 MYTHS ABOUT AUTOMATION, ALGORITHMS, SOCIETY AND OURSELVES25
What AI scientists and practicioners can do
❖ Sociotechnical embedding: AI & related technologies
❖ Democratization: Inclusion of all stakeholders, especially the persons affected by the AI
❖ Design: Participatory and inclusive
❖ Public Ownership: Data & digital infrastructures
❖ Open Science: Transparency & comprehensibility
❖ Self-Reflection: on your situating in the world
What we can do
❖ Let’s educate ourselves and enhance our digital literacy (like you do, right now!)
❖ Support algorithmic watchdogs
❖ Have a public debate around the areas in which AI should be deployed.
There is no technical fix for discrimination.
Fight or support fights against discrimination in all parts of society.
WHY AI – UNRAVELING 15 MYTHS ABOUT AUTOMATION, ALGORITHMS, SOCIETY AND OURSELVES26
Phillip Lücking Miriam Fahimi [email protected] [email protected]
Thank you.